Multi-view learning-based data proliferator for boosting classification using highly imbalanced classes.

Area of Science:

• Neuroscience
• Computer Science
• Medical Imaging

Background:

• Multi-view data representation learning enhances computer-aided diagnosis by leveraging complementary information across different data sources.
• Machine learning for neurological disorder diagnosis, such as autism spectrum disorder (ASD), faces challenges with imbalanced and non-heterogeneous multi-view brain data.
• Collecting large, balanced medical datasets for training is difficult and expensive in clinical settings.

Purpose of the Study:

• To address the unexplored problem of imbalanced and multi-view data classification in network neuroscience.
• To propose a novel method, Multi-View LEArning-based data Proliferator (MV-LEAP), for classifying imbalanced multi-view representations.
• To improve the accuracy of automated diagnosis for neurological disorders using challenging clinical datasets.

Main Methods:

• MV-LEAP utilizes a manifold learning-based proliferator to generate synthetic data, effectively handling imbalanced classes.
• It employs multi-view manifold data alignment through tensor canonical correlation analysis to harmonize data distributions.
• All original and synthesized views are mapped into a shared subspace for improved classification.

Main Results:

• The method was evaluated on imbalanced multi-view datasets comparing autism spectrum disorder (ASD) patients with normal controls (NC).
• MV-LEAP demonstrated superior performance in classifying these challenging datasets.
• The proposed approach outperformed existing baseline data synthesis methods.

Conclusions:

• MV-LEAP successfully addresses the critical issue of imbalanced and multi-view data in network neuroscience.
• The method provides a robust solution for computer-aided diagnosis in the presence of data limitations.
• This work advances the field by enabling more accurate classification of neurological disorders from complex brain data.